CN113627279A - High-voltage circuit breaker fault diagnosis method based on support vector machine - Google Patents
High-voltage circuit breaker fault diagnosis method based on support vector machine Download PDFInfo
- Publication number
- CN113627279A CN113627279A CN202110829807.6A CN202110829807A CN113627279A CN 113627279 A CN113627279 A CN 113627279A CN 202110829807 A CN202110829807 A CN 202110829807A CN 113627279 A CN113627279 A CN 113627279A
- Authority
- CN
- China
- Prior art keywords
- current
- circuit breaker
- support vector
- feature
- vector machine
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 41
- 238000003745 diagnosis Methods 0.000 title claims abstract description 14
- 238000000513 principal component analysis Methods 0.000 claims abstract description 39
- 239000013598 vector Substances 0.000 claims abstract description 36
- 230000007246 mechanism Effects 0.000 claims abstract description 21
- 230000005540 biological transmission Effects 0.000 claims abstract description 19
- 238000012360 testing method Methods 0.000 claims abstract description 12
- 238000004088 simulation Methods 0.000 claims abstract description 11
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 8
- 230000009467 reduction Effects 0.000 claims description 27
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical group [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims description 17
- 239000011159 matrix material Substances 0.000 claims description 16
- 230000001186 cumulative effect Effects 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 10
- 230000007423 decrease Effects 0.000 claims description 9
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000012567 pattern recognition method Methods 0.000 claims description 4
- 239000006185 dispersion Substances 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000005259 measurement Methods 0.000 claims description 3
- 238000011946 reduction process Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 abstract description 3
- 230000006872 improvement Effects 0.000 description 8
- 230000000694 effects Effects 0.000 description 4
- 230000033001 locomotion Effects 0.000 description 4
- 238000012545 processing Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000011425 standardization method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Signal Processing (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Arc-Extinguishing Devices That Are Switches (AREA)
Abstract
The invention discloses a high-voltage circuit breaker fault diagnosis method based on a support vector machine, which comprises the following steps of: step S1, adding 4 characteristics of current mean value, standard deviation, kurtosis and energy parameter on the basis of local characteristics, and using the characteristics as the global characteristics of current signals; then, the local features and the global features are used as feature vectors of current signals, and partial fault states of the circuit breaker are identified; step S2, evaluating the running state of the circuit breaker by using current time, current value, current mean value, standard deviation, kurtosis and energy as characteristic parameters and adopting an algorithm combining principal component analysis and a support vector machine; and step S3, performing simulation tests on 4 faults of normal state, fatigue of a closing spring, looseness of a transmission mechanism and low control loop voltage. The invention adopts the PCA-SVM method to evaluate the operating state of the circuit breaker, can quickly extract effective information of signal characteristics and effectively identify the operating state of the circuit breaker.
Description
Technical Field
The invention relates to a high-voltage circuit breaker fault diagnosis method based on a support vector machine, and belongs to the technical field of high-voltage circuit breaker diagnosis.
Background
The high-voltage circuit breaker plays an important role in protection and control in a power transmission and distribution network, and the high fault rate is caused by the severe working environment, so that the fault diagnosis of the circuit breaker is very necessary. The opening and closing coil current signals can reflect the working conditions of the iron core jamming of the circuit breaker, the jamming of a transmission mechanism, the voltage of a control loop and the like, and meanwhile, the circuit breaker has non-invasiveness in collection, and the normal operation of the circuit breaker cannot be influenced.
In the previous research on current signals, the current time and the current value of the opening and closing coil are effectively extracted on the basis of a fuzzy theory, and a basis is provided for the subsequent analysis of the fault state of the circuit breaker. A method based on spline interpolation and multi-scale linear fitting is used for extracting characteristics of currents of opening and closing coils of a circuit breaker and can accurately distinguish different states of the circuit breaker. The current time and the current value are taken as characteristics, the mean value and the variance of the characteristic quantity are calculated, the 95% confidence interval is taken as a threshold value, a fault diagnosis system is constructed, and the fault current detection result is good. But the above studies only consider local features in the time domain.
When the circuit breaker simulated fault signal is analyzed, considering that the correlation of various characteristic parameters can cause difficulty in analysis, and the more characteristic parameters can cause long classification training time and influence the recognition effect, an effective data processing method is needed to reduce the data calculation amount and quickly extract effective signal information. In common pattern recognition, most methods need enough training samples to ensure the classification accuracy, and the number of breaker simulation fault samples is limited, so that a pattern recognition method suitable for small sample classification needs to be selected.
Disclosure of Invention
The invention aims to solve the technical problem of providing a high-voltage circuit breaker fault diagnosis method based on a support vector machine, which is characterized in that the operating state of the circuit breaker is evaluated by adopting a PCA-SVM method by taking current time, current value, current mean value, standard deviation, kurtosis and energy as characteristic parameters, so that effective information of signal characteristics can be quickly extracted and the operating state of the circuit breaker can be effectively identified.
In order to solve the problems, the technical scheme adopted by the invention is as follows:
a high-voltage circuit breaker fault diagnosis method based on a support vector machine comprises the following steps:
step S1, adding 4 characteristics of current mean value, standard deviation, kurtosis and energy parameter on the basis of local characteristics, and using the characteristics as the global characteristics of current signals; then, the local features and the global features are used as feature vectors of current signals, and partial fault states of the circuit breaker are identified;
step S2, evaluating the running state of the circuit breaker by using current time, current value, current mean value, standard deviation, kurtosis and energy as characteristic parameters and adopting an algorithm combining principal component analysis and a support vector machine;
and step S3, performing simulation tests on 4 faults of normal state, fatigue of a closing spring, looseness of a transmission mechanism and low control loop voltage.
As a further improvement of the present invention, in step S1, the local characteristics include a current time and a current value;
in step S2, the principal component analysis is a method of converting high-dimensional space information into low-dimensional space by coordinate transformation, which is called PCA method; the support vector machine is a pattern recognition method which minimizes the error of a sample point and reduces the upper bound of the model prediction error on the basis of limited sample information, and is called as an SVM method.
As a further improvement of the present invention, the current mean value reflects the stability of the signal, and is calculated by the following formula:
where μ is the current mean value, xiThe value of the random variable is obtained;
the standard deviation describes the degree of dispersion of the data, and is used to characterize the energy of the signal in signal analysis, and is calculated by the following formula:
wherein σ is a standard deviation;
the kurtosis is a fourth order statistic which reflects the signal distribution characteristic and is the simplest measurement of non-Gaussian of a random variable, and the calculation method is as follows:
wherein K is kurtosis;
the energy represents the working magnitude of the closing current, the magnitude of the energy depends on the voltage, the current and the time, and the calculation method is as follows:
in the formula, W is energy, U is control loop voltage, I is closing coil current value, and T is time.
As a further improvement of the invention, the algorithm flow of the principal component analysis combined with the support vector machine comprises the following steps:
step S21, loading data;
step S22, feature extraction, wherein current time, current value, current mean value, standard deviation, kurtosis and energy feature parameters are extracted;
step S23, PCA feature dimension reduction, which is to standardize all features and carry out PCA dimension reduction to obtain effective feature vectors; step S24, SVM feature classification, inputting the obtained effective feature vector into SVM for training and prediction to obtain a working state classification result;
and step S25, calculating the training time and the classification accuracy.
As a further improvement of the invention, in the operation process of the high-voltage circuit breaker, the current signal of the switching-on/off coil is divided into 4 stages:
0-T1: at the time 0, the coil starts to be electrified, the current rises exponentially, and in the process, the electromagnetic force generated by the current is gradually enhanced; to T1The current reaches the first wave crest at the moment, the electromagnetic force generated by the current is larger than the external resistance borne by the iron core, and the iron core starts to move; the condition of coil voltage, loop resistance, whether the iron core is clamped or not and the like can be reflected in the stage;
T1-T2：T1thereafter, the core begins to move, whereupon the coil current decreases, to T2At that moment, the current reaches a minimum value. During this process the core speed decreases sharply to T2At that moment, the iron core impacts the locking/tripping device and then stops moving; the stage can reflect faults such as jamming of the iron core movement and failure of the notch;
T2-T3：T2at the moment, after the iron core stops moving, the lock catch of the transmission mechanism is opened, the current continuously rises at the stage, and the moving contact starts to actThe opening spring starts opening; to T3At that time, the current reaches a peak value; the stage can reflect the motion state of the transmission mechanism;
T3-T4：T3at the moment, the movable contact and the fixed contact are completely separated, the auxiliary switch is switched off, the coil power supply is cut off, electric arcs are generated between the contacts, the voltage of the electric arcs is sharply increased in a short time, and the current is directly sharply reduced; t is4At that time, the current decreases to 0. The current profile at this stage may reflect whether the auxiliary switch has a fault.
As a further development of the invention, the current time includes four times T1、T2、T3And T4(ii) a The current values include respectively corresponding to T1、T2And T3Opening and closing coil current I at moment1、I2And I3。
As a further improvement of the invention, the PCA feature dimension reduction process is as follows: suppose that the vibration signal of the circuit breaker has k training samples, each sample has n characteristic values, and the column vector of the sample is xk＝(x1k，x2k，…，xnk)TThe average vector of the sample set is
Covariance matrix
Determining all eigenvalues lambda of the covariance matrixiAnd a feature vector viThe eigenvectors form an m-dimensional orthogonal space; the characteristic value lambda is divided into a large characteristic value and a small characteristic value in the order from large to smalliArranged as lambda1＞λ2＞…＞λd＞λd+1More than …, sample information is concentrated in eigenvectors with larger eigenvalues, and discarding some eigenvectors with small eigenvalues does not affect the information, so that the selection of eigenvectors larger than lambdadThe feature vectors of (a) constitute principal components; sample xiProjection onto feature vector viObtaining a principal component corresponding to the direction
Let d principal components in orthogonal space be y1，y2，…，ydThen the cumulative variance contribution rate is
The cumulative variance contribution rate determines the selection number of the principal components, and the cumulative variance contribution rate is H (d) more than 95%, namely only the first d principal components are selected to represent the original information, and the first d principal components contain more than 95% of effective information of the original signal, thereby achieving the purpose of reducing the dimension.
As a further improvement of the present invention, in step S3, the simulation test follows a single variable principle, and there is only one fault at a time; the fatigue of the closing spring is realized by loosening a fixing bolt of the closing spring, and the loosening of the transmission mechanism is realized by adjusting the length of a connecting rod of the mechanism.
As a further improvement of the invention, in the simulation test, the current time and the current value adopt a 3-time spline interpolation method to carry out smooth processing on the original signal to obtain four current times T1、T2、T3And T4，And three current values I1、I2And I3(ii) a Obtaining the characteristics of mean value mu, standard deviation sigma, kurtosis K and energy W of the current signal by using the formulas (1) - (4);
selecting closing current signals of 4 states of a circuit breaker in a normal state, closing spring fatigue, transmission mechanism looseness and control loop voltage low, and acquiring 30 groups of data in each state to extract characteristic parameters to form a 120 multiplied by 11 characteristic matrix. As a further improvement of the invention, the original local features, the local features after PCA dimension reduction, the original comprehensive features and the comprehensive features after PCA dimension reduction are classified and identified by an SVM respectively; and comparing the original local features with the classification recognition results of the local features after the PCA dimension reduction, and comparing the original comprehensive features with the classification recognition results of the comprehensive features after the PCA dimension reduction so as to judge the advantages of the feature vectors after the PCA dimension reduction in the classification accuracy and the program running time.
Adopt the produced beneficial effect of above-mentioned technical scheme to lie in:
the invention fits the fatigue of a brake spring, the looseness of a transmission mechanism and the low voltage of a control loop on a breaker in 3 fault states, extracts local and global characteristic parameters of current signals in 4 states of the breaker by combining current signals in a normal state of the breaker, adopts a PCA method to reduce the dimension of signal characteristics, inputs the feature vectors after the dimension reduction into an SVM for classification, and shows that:
(1) on the basis of the traditional characteristic of current time and current value, 4 global characteristics of mean value, standard deviation, kurtosis and energy are added, so that the classification accuracy is improved, and the operation condition of the circuit breaker can be more accurately reflected.
(2) The feature vector after the dimensionality reduction is adopted, the classification accuracy is improved compared with the original feature vector, the training time is reduced, the PCA method can retain effective feature information of signals, interference information is eliminated, and the calculation efficiency is improved.
(3) The comprehensive classification performance of the SVM on the small sample classification problem is superior to that of a BP neural network, and the SVM has good performance in fault diagnosis of the circuit breaker.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed for the embodiments or the prior art descriptions will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a schematic view of a typical opening and closing coil current curve;
FIG. 2 is a flow chart of a fault diagnosis algorithm based on PCA and SVM;
FIG. 3 is a graph showing the variation trend of characteristic parameters of a closing current signal;
FIG. 4 is a table of power characteristic parameters of the closing coil;
FIG. 5 is a single principal component contribution rate and cumulative contribution rate table;
FIG. 6 is a histogram of individual principal component contribution rates versus cumulative contribution rates;
FIG. 7 is a partial feature parameter table after PCA dimensionality reduction;
FIG. 8 is a diagram of SVM classification results;
FIG. 9 is a comparison table of the classification effect of different features.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the application, its application, or uses. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments according to the present application. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
The relative arrangement of the components and steps, the numerical expressions, and numerical values set forth in these embodiments do not limit the scope of the present application unless specifically stated otherwise. Techniques, methods, and apparatus known to those of ordinary skill in the relevant art may not be discussed in detail but are intended to be part of the specification where appropriate. In all examples shown and discussed herein, any particular value should be construed as merely illustrative, and not limiting.
The embodiment provides a breaker fault diagnosis method based on opening and closing coil current. Firstly, extracting current time and current value of a circuit breaker opening and closing coil, and taking the current time and current value as local characteristics; then, calculating the average value, the standard deviation, the kurtosis and the energy value of the current signal data, taking the values as global characteristics, and forming comprehensive characteristics by the local characteristics and the global characteristics; then, carrying out dimensionality reduction processing on the feature matrix by utilizing a PCA (principal component analysis) method to obtain principal component features; and finally, constructing a training sample set of the SVM (support vector machine), obtaining an SVM decision function, and realizing the classification of the operation state of the circuit breaker. The method is used for carrying out test analysis on 4 operating states of the circuit breaker, and the result shows that the method can effectively identify different fault states of the circuit breaker.
A breaker fault diagnosis method based on opening and closing coil current comprises the following steps:
step S1, adding 4 characteristics of current mean value, standard deviation, kurtosis and energy parameter on the basis of local characteristics, and using the characteristics as the global characteristics of current signals; then, the local features and the global features are used as feature vectors of current signals, and partial fault states of the circuit breaker are identified;
step S2, evaluating the running state of the circuit breaker by using current time, current value, current mean value, standard deviation, kurtosis and energy as characteristic parameters and adopting an algorithm combining principal component analysis and a support vector machine;
and step S3, performing simulation tests on 4 faults of normal state, fatigue of a closing spring, looseness of a transmission mechanism and low control loop voltage.
Specifically, in step S1, the local characteristics include a current time and a current value;
in step S2, the principal component analysis is a method of converting high-dimensional space information into low-dimensional space by coordinate transformation, which is called PCA method; the support vector machine is a pattern recognition method which minimizes the error of a sample point and reduces the upper bound of the model prediction error on the basis of limited sample information, and is called as an SVM method.
In this embodiment, the current mean reflects the stability of the signal, and the calculation method is as follows:
where μ is the current mean value, xiThe value of the random variable is obtained;
the standard deviation describes the degree of dispersion of the data, and is used to characterize the energy of the signal in signal analysis, and is calculated by the following formula:
wherein σ is a standard deviation;
the kurtosis is a fourth order statistic which reflects the signal distribution characteristic and is the simplest measurement of non-Gaussian of a random variable, and the calculation method is as follows:
wherein K is kurtosis;
the energy represents the working magnitude of the closing current, the magnitude of the energy depends on the voltage, the current and the time, and the calculation method is as follows:
in the formula, W is energy, U is control loop voltage, I is closing coil current value, and T is time.
As shown in fig. 2, in this embodiment, the algorithm flow of the principal component analysis combined with the support vector machine includes the following steps:
step S21, loading data;
step S22, feature extraction, wherein current time, current value, current mean value, standard deviation, kurtosis and energy feature parameters are extracted;
step S23, PCA feature dimension reduction, which is to standardize all features and carry out PCA dimension reduction to obtain effective feature vectors;
step S24, SVM feature classification, inputting the obtained effective feature vector into SVM for training and prediction to obtain a working state classification result;
and step S25, calculating the training time and the classification accuracy.
As shown in fig. 1, in this embodiment, specifically, in the operation process of the high-voltage circuit breaker, the opening/closing coil current signal is divided into 4 stages:
0-T1: at the time 0, the coil starts to be electrified, the current rises exponentially, and in the process, the electromagnetic force generated by the current is gradually enhanced; to T1The current reaches the first wave crest at the moment, the electromagnetic force generated by the current is larger than the external resistance borne by the iron core, and the iron core starts to move; the condition of coil voltage, loop resistance, whether the iron core is clamped or not and the like can be reflected in the stage;
T1-T2：T1thereafter, the core begins to move, whereupon the coil current decreases, to T2At that moment, the current reaches a minimum value. During this process the core speed decreases sharply to T2At that moment, the iron core impacts the locking/tripping device and then stops moving; the stage can reflect faults such as jamming of the iron core movement and failure of the notch;
T2-T3：T2at the moment, after the iron core stops moving, the lock catch of the transmission mechanism is opened, the current continuously rises at the stage, the moving contact starts to act, and the opening spring starts to open; to T3At that time, the current reaches a peak value; the stage can reflect the motion state of the transmission mechanism;
T3-T4：T3at the moment, the movable contact and the fixed contact are completely separated, the auxiliary switch is switched off, the coil power supply is cut off, electric arcs are generated between the contacts, the voltage of the electric arcs is sharply increased in a short time, and the current is directly sharply reduced; t is4At that time, the current decreases to 0. The current profile at this stage may reflect whether the auxiliary switch has a fault.
In this embodiment, the current time includes four moments T1、T2、T3And T4(ii) a The current values include respectively corresponding to T1、T2And T3Opening and closing coil current I at moment1、I2And I3。
Specifically, in this embodiment, the PCA feature dimension reduction process is as follows: suppose that the vibration signal of the circuit breaker has k training samples, each sample has n characteristic values, and the column vector of the sample is xk＝(x1k，x2k，…，xnk)TThe average vector of the sample set is
Covariance matrix
Determining all eigenvalues lambda of the covariance matrixiAnd a feature vector viThe eigenvectors form an m-dimensional orthogonal space; the characteristic value lambda is divided into a large characteristic value and a small characteristic value in the order from large to smalliArranged as lambda1＞λ2＞…＞λd＞λd+1More than …, sample information is concentrated in eigenvectors with larger eigenvalues, and discarding some eigenvectors with small eigenvalues does not affect the information, so that the selection of eigenvectors larger than lambdadThe feature vectors of (a) constitute principal components; sample xiProjection onto feature vector viObtaining a principal component corresponding to the direction
Let d principal components in orthogonal space be y1，y2，…，ydThen the cumulative variance contribution rate is
The cumulative variance contribution rate determines the selection number of the principal components, and the cumulative variance contribution rate is H (d) more than 95%, namely only the first d principal components are selected to represent the original information, and the first d principal components contain more than 95% of effective information of the original signal, thereby achieving the purpose of reducing the dimension.
In this embodiment, in step S3, the simulation test follows a single variable principle, and there is only one fault in each circuit breaker; the fatigue of the closing spring is realized by loosening a fixing bolt of the closing spring, and the loosening of the transmission mechanism is realized by adjusting the length of a connecting rod of the mechanism.
In this embodiment, in the simulation test, the current time and the current value are obtained by smoothing the original signal by using a 3-time spline interpolation method to obtain four current times T1、T2、T3And T4，And three current values I1、I2And I3(ii) a Obtaining the characteristics of mean value mu, standard deviation sigma, kurtosis K and energy W of the current signal by using the formulas (1) - (4);
selecting closing current signals of 4 states of a circuit breaker in a normal state, fatigue of a closing spring, looseness of a transmission mechanism and low control loop voltage, and acquiring 30 groups of data in each state to extract characteristic parameters to form a 120 multiplied by 11 characteristic matrix, wherein part of data is shown in figure 4.
Because the physical meanings represented by each characteristic parameter are different and the sizes of the parameters are different, the parameters are standardized by adopting a mean-variance standardization method. In fig. 3, the number of sample points is used as the abscissa, and the normalized value is used as the ordinate, which reflects the variation trend of 11 characteristic parameters of the closing current signal in 4 states.
As can be seen from fig. 3, each characteristic parameter has different reflection trends and regularity for the operating state of the circuit breaker, and some parameters insensitive to the fault state increase the calculation load and should be filtered out, so the PCA method is used to perform the dimension reduction processing on the characteristic parameter matrix to extract the effective information of the characteristic parameters.
The signals were processed using the PCA method as follows: all signals were normalized by mean-variance to eliminate differences between different dimensions. And solving a covariance matrix of the characteristic parameter matrix. And solving the eigenvalue and the eigenvector of the covariance matrix, and arranging the eigenvalues in the order from big to small to obtain the corresponding eigenvector. And calculating principal components and principal component contribution rates. The principal components are feature vectors arranged in sequence, the contribution rate of the principal components is the proportion of effective information contained in each principal component, and the contribution rate of the principal components is the feature value/sigma feature value.
Fig. 5 shows 11 principal component contribution rates and the cumulative contribution rate of the principal components, which are plotted as curves as shown in fig. 6, so that the laws and features of the individual principal component contribution rates and the cumulative contribution rate of the principal components can be more intuitively represented. The PCA essentially takes the direction with the largest variance as a main feature, the larger the variance is, the more obvious the change in the direction is, and the larger the contribution rate is in the vector reconstruction process, so the reconstruction components are arranged in the order of the variance from large to small, and the first N principal components with the cumulative contribution rate not less than 95% are selected. As can be seen from fig. 6, the cumulative contribution rate of the first 3 principal components reaches 97.84%, so the first 3 principal components are selected as new feature vectors to characterize the original signal. Part of the principal component features are shown in fig. 8.
30 groups of closing coil current signals of 4 states of a breaker normal state, closing spring fatigue, transmission mechanism looseness and control loop voltage low are taken, and the dimension reduction is carried out by PCA to obtain a 120 multiplied by 6 feature matrix, wherein each state corresponds to a 30 multiplied by 6 feature matrix. The first 24 groups of data are selected for training, the last 6 groups of data are selected for testing, the radial basis kernel function is selected for carrying out support vector classification training on the samples, 4 one-to-three SVM are constructed, the obtained classification result is shown in FIG. 7, the accuracy is 91.67%, and the effect is good.
In this embodiment, the original local features, the local features after the PCA dimension reduction, the original comprehensive features, and the comprehensive features after the PCA dimension reduction are classified and identified by using an SVM, and the obtained result is shown in fig. 9; the original local features are compared with the classification recognition results of the local features after the PCA dimension reduction, the original comprehensive features are compared with the classification recognition results of the comprehensive features after the PCA dimension reduction, so that the feature vector after the PCA dimension reduction has obvious advantages in classification accuracy and program running time, the PCA can be explained to remove redundant components on the basis of ensuring the effective information of signals, and the calculation efficiency is improved; global features are added on the basis of the traditional current feature vectors to form comprehensive features, so that the circuit breaker state identification effect is better, and the classification accuracy is effectively improved.
Claims (10)
1. A high-voltage circuit breaker fault diagnosis method based on a support vector machine is characterized by comprising the following steps:
step S1, adding 4 characteristics of current mean value, standard deviation, kurtosis and energy parameter on the basis of local characteristics, and using the characteristics as the global characteristics of current signals; then, the local features and the global features are used as feature vectors of current signals, and partial fault states of the circuit breaker are identified;
step S2, evaluating the running state of the circuit breaker by using current time, current value, current mean value, standard deviation, kurtosis and energy as characteristic parameters and adopting an algorithm combining principal component analysis and a support vector machine;
and step S3, performing simulation tests on 4 faults of normal state, fatigue of a closing spring, looseness of a transmission mechanism and low control loop voltage.
2. The method for diagnosing the fault of the high-voltage circuit breaker based on the support vector machine according to claim 1,
in step S1, the local features include current time and current value;
in step S2, the principal component analysis is a method of converting high-dimensional space information into low-dimensional space by coordinate transformation, which is called PCA method; the support vector machine is a pattern recognition method which minimizes the error of a sample point and reduces the upper bound of the model prediction error on the basis of limited sample information, and is called as an SVM method.
3. The method as claimed in claim 2, wherein the current mean value reflects a stability of the signal, and is calculated as follows:
where μ is the current mean value, xiThe value of the random variable is obtained;
the standard deviation describes the degree of dispersion of the data, and is used to characterize the energy of the signal in signal analysis, and is calculated by the following formula:
wherein σ is a standard deviation;
the kurtosis is a fourth order statistic which reflects the signal distribution characteristic and is the simplest measurement of non-Gaussian of a random variable, and the calculation method is as follows:
wherein K is kurtosis;
the energy represents the working magnitude of the closing current, the magnitude of the energy depends on the voltage, the current and the time, and the calculation method is as follows:
in the formula, W is energy, U is control loop voltage, I is closing coil current value, and T is time.
4. The method for diagnosing the fault of the high-voltage circuit breaker based on the support vector machine as recited in claim 3, wherein the algorithm flow of the principal component analysis combined with the support vector machine comprises the following steps:
step S21, loading data;
step S22, feature extraction, wherein current time, current value, current mean value, standard deviation, kurtosis and energy feature parameters are extracted;
step S23, PCA feature dimension reduction, which is to standardize all features and carry out PCA dimension reduction to obtain effective feature vectors;
step S24, SVM feature classification, inputting the obtained effective feature vector into SVM for training and prediction to obtain a working state classification result;
and step S25, calculating the training time and the classification accuracy.
5. The method for diagnosing the fault of the high-voltage circuit breaker based on the support vector machine according to claim 4, wherein in the operation process of the high-voltage circuit breaker, the current signal of the opening and closing coil is divided into 4 stages:
0-T1: at the time 0, the coil starts to be electrified, the current rises exponentially, and in the process, the electromagnetic force generated by the current is gradually enhanced; to T1The current reaches the first wave crest at the moment, the electromagnetic force generated by the current is larger than the external resistance borne by the iron core, and the iron core starts to move;
T1-T2：T1thereafter, the core begins to move, whereupon the coil current decreases, to T2At that moment, the current reaches a minimum value. During this process the core speed decreases sharply to T2At that moment, the iron core impacts the locking/tripping device and then stops moving;
T2-T3：T2at the moment, after the iron core stops moving, the lock catch of the transmission mechanism is opened, the current continuously rises at the stage, the moving contact starts to act, and the opening spring starts to open; to T3At that time, the current reaches a peak value;
T3-T4：T3at the moment, the movable contact and the fixed contact are completely separated, the auxiliary switch is switched off, the coil power supply is cut off, electric arcs are generated between the contacts, the voltage of the electric arcs is sharply increased in a short time, and the current is directly sharply reduced; t is4At that time, the current decreases to 0.
6. The method as claimed in claim 5, wherein the current time includes four times T, T1、T2、T3And T4(ii) a The current values include respectively corresponding to T1、T2And T3Opening and closing coil current I at moment1、I2And I3。
7. The method for diagnosing the fault of the high-voltage circuit breaker based on the support vector machine as claimed in claim 6, wherein the PCA characteristic dimension reduction process is as follows: suppose that the vibration signal of the circuit breaker has k training samples, each sample has n characteristic values, and the column vector of the sample is xk＝(x1k，x2k，…，xnk)TThe average vector of the sample set is
Covariance matrix
Determining all eigenvalues lambda of the covariance matrixiAnd a feature vector viThe eigenvectors form an m-dimensional orthogonal space; the characteristic value lambda is divided into a large characteristic value and a small characteristic value in the order from large to smalliArranged as lambda1＞λ2＞…＞λd＞λd+1More than …, sample information is concentrated in eigenvectors with larger eigenvalues, and discarding some eigenvectors with small eigenvalues does not affect the information, so that the selection of eigenvectors larger than lambdadThe feature vectors of (a) constitute principal components; sample xiProjection onto feature vector viObtaining a principal component corresponding to the direction
Let d principal components in orthogonal space be y1，y2，…，ydThen the cumulative variance contribution rate is
The cumulative variance contribution rate determines the selection number of the principal components, and the cumulative variance contribution rate is H (d) more than 95%, namely only the first d principal components are selected to represent the original information, and the first d principal components contain more than 95% of effective information of the original signal, thereby achieving the purpose of reducing the dimension.
8. The method for diagnosing the fault of the high-voltage circuit breaker based on the support vector machine as recited in claim 7, wherein in the step S3, the simulation test follows a single variable principle, and only one fault exists in the circuit breaker at a time; the fatigue of the closing spring is realized by loosening a fixing bolt of the closing spring, and the loosening of the transmission mechanism is realized by adjusting the length of a connecting rod of the mechanism.
9. The method for diagnosing the faults of the high-voltage circuit breaker based on the support vector machine as claimed in claim 8, wherein during the simulation test, the current time and the current value are smoothed by adopting a 3-time spline interpolation method to obtain four current times T1、T2、T3And T4，And three current values I1、I2And I3(ii) a Obtaining the characteristics of mean value mu, standard deviation sigma, kurtosis K and energy W of the current signal by using the formulas (1) - (4);
selecting closing current signals of 4 states of a circuit breaker in a normal state, closing spring fatigue, transmission mechanism looseness and control loop voltage low, and acquiring 30 groups of data in each state to extract characteristic parameters to form a 120 multiplied by 11 characteristic matrix.
10. The method for diagnosing the fault of the high-voltage circuit breaker based on the support vector machine as recited in claim 9, wherein the original local feature, the local feature after PCA dimension reduction, the original comprehensive feature and the comprehensive feature after PCA dimension reduction are classified and identified by SVM respectively; and comparing the original local features with the classification recognition results of the local features after the PCA dimension reduction, and comparing the original comprehensive features with the classification recognition results of the comprehensive features after the PCA dimension reduction so as to judge the advantages of the feature vectors after the PCA dimension reduction in the classification accuracy and the program running time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110829807.6A CN113627279A (en) | 2021-07-22 | 2021-07-22 | High-voltage circuit breaker fault diagnosis method based on support vector machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110829807.6A CN113627279A (en) | 2021-07-22 | 2021-07-22 | High-voltage circuit breaker fault diagnosis method based on support vector machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113627279A true CN113627279A (en) | 2021-11-09 |
Family
ID=78380492
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110829807.6A Pending CN113627279A (en) | 2021-07-22 | 2021-07-22 | High-voltage circuit breaker fault diagnosis method based on support vector machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113627279A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114757110A (en) * | 2022-06-16 | 2022-07-15 | 南昌航空大学 | Circuit breaker fault diagnosis method based on sliding window detection and current extraction signals |
-
2021
- 2021-07-22 CN CN202110829807.6A patent/CN113627279A/en active Pending
Non-Patent Citations (1)
| Title |
|---|
| 周建平,李聪,万书亭等: "基于优化型SVM的高压断路器故障诊断方法研究", 浙江电力 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114757110A (en) * | 2022-06-16 | 2022-07-15 | 南昌航空大学 | Circuit breaker fault diagnosis method based on sliding window detection and current extraction signals |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110221200B (en) | Universal circuit breaker accessory fault diagnosis method based on deep learning | |
| CN106199412B (en) | A kind of permanent magnet mechanism high-pressure vacuum breaker method of fault pattern recognition | |
| CN107101813B (en) | A kind of frame-type circuit breaker mechanical breakdown degree assessment method based on vibration signal | |
| CN106017879B (en) | Omnipotent breaker mechanical failure diagnostic method based on acoustic signal Fusion Features | |
| CN112200032A (en) | Attention mechanism-based high-voltage circuit breaker mechanical state online monitoring method | |
| CN105891707A (en) | Opening-closing fault diagnosis method for air circuit breaker based on vibration signals | |
| CN105259495A (en) | High-voltage circuit breaker operation mechanism state evaluation method based on opening-closing coil current characteristic quantity optimization | |
| CN112345213A (en) | Low-voltage direct-current circuit breaker mechanical fault diagnosis method | |
| CN113934982B (en) | Method for predicting mechanical life of breaker operating mechanism based on vibration-electric signal fusion | |
| CN108919104B (en) | Breaker fault diagnosis method based on Fisher discriminant classification method | |
| CN113627672A (en) | Circuit breaker dynamic contact resistance prediction method based on partial least square estimation | |
| CN113191192A (en) | Breaker fault detection method based on wavelet analysis and fuzzy neural network algorithm | |
| CN112684329A (en) | Intelligent diagnosis method for mechanical fault of high-voltage circuit breaker | |
| CN113627279A (en) | High-voltage circuit breaker fault diagnosis method based on support vector machine | |
| CN114757110B (en) | Circuit breaker fault diagnosis method based on sliding window detection and current extraction signals | |
| CN117691756B (en) | Safety early warning management method and system for power distribution cabinet | |
| CN115113038B (en) | Circuit breaker fault diagnosis method based on current signal phase space reconstruction | |
| CN112698194A (en) | Comprehensive evaluation method and system for state of circuit breaker operating mechanism | |
| CN112363050A (en) | SF6 circuit breaker arc contact state evaluation method based on dynamic contact resistance signal | |
| CN116125260A (en) | Breaker electromechanical fault edge diagnosis method based on multi-element data fusion | |
| CN116990633A (en) | Fault studying and judging method based on multiple characteristic quantities | |
| CN110824344A (en) | High-voltage circuit breaker state evaluation method based on vibration signal short-time energy-entropy ratio and DTW | |
| CN115728628A (en) | On-line monitoring and early warning method and device for circuit breaker control loop fault | |
| CN115238733A (en) | Method for evaluating operation state of switching-on and switching-off coil of high-voltage circuit breaker and related equipment | |
| CN114839523A (en) | Circuit breaker fault detection system and method based on CNN combined LSTM algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211109 |